Silent output audit proof.

A mock proof for n8n or Make workflows that technically succeed but still do the wrong thing. Synthetic run rows are checked for wrong tools, wrong context, bad tone, broken metrics, and client-visible mistakes before any live workflow, customer message, or client-visible dashboard is touched.

Internal Audit Ledger CSV Client Status View CSV Review Queue CSV Blocked Action Queue CSV Error Log CSV Sample Runs CSV Expected Output Rules CSV Runbook Source Script

Input

Ten synthetic workflow runs across appointment agents, support summaries, quote builders, status digests, review replies, and invoice replies.

Checks

Expected tool/action, context key, output tone, business metric range, alert state, client visibility, source approval, sensitivity, and duplicate key.

Output

Internal audit ledger, client-safe status view, review queue, blocked-action queue, hard error log, source script, and written runbook.

Run	Workflow	Decision	Why	Boundary
run-9101	Renewal Risk Digest	Accepted	Expected action, context, tone, and metric checks passed.	Safe summary can appear in the client status view.
run-9103	Support Summary Bot	Review	Workflow looked successful but summarized the wrong customer context.	Review with redacted evidence before changing production behavior.
run-9102	Appointment Reschedule Agent	Blocked	Wrong action reached a client-visible path.	Hold before any customer message or client-visible publication.
run-9108	Lead Enrichment Agent	Hard Error	Required client context is missing.	Fail visibly instead of guessing workflow ownership.